The No Labels problem
Computational simulations can test whether they make a Trump win less likely - or more likely.
Today’s Supreme Court decision removes courts from determining whether Donald Trump’s insurrectionism is disqualifying from the ballot. Tomorrow is Super Tuesday, which will continue his inevitable march to the Republican nomination. So the main arbiter of whether Donald Trump returns to office will be voters in November. If that is an outcome you want to stop, then the question is what actions materially move that probability - in either direction. Let’s consider the No Labels organization, and ask: which way do they move the Presidential race?
In past months I have written about ways to achieve practical change. Today I will dive deeper into the word “practical,” in the context of the No Labels organization.
For a repair to democracy to be worth pursuing, it’s not enough to be conceptually appealing. The repair also has to achieve the effects that its proponents want. Indeed, those practical effects are more important than conceptual appeal. Good intentions have limited value if they lead to bad outcomes. And this is where No Labels in danger of going off track.
No Labels seeks to find a middle way between the major parties. Toward that goal, they have taken a variety of approaches. They’re in the news because one stratagem, a third-party “unity” Presidential ticket, has run into strong headwinds, or according to Greg Sargent at the New Republic, a meltdown.
In the past, No Labels has considered other paths to improving democracy. They formed the U.S. House Problem Solvers Caucus, a bipartisan group of Democratic and Republican lawmakers who have nominally committed to working together. And recently, one of their founders, journalist John Avlon, is running for Congress. Will any of these approaches work?
Proponents of the Unity ticket concept point out that the two parties have drifted far apart in policy positions, and that deep partisanship prevents sensible work from getting done in Congress. However, they also recognize the considerable threat to democracy posed by Donald Trump. Critics of No Labels, such as Third Way, point out that Donald Trump is the one person who a Unity ticket might aid the most. They and retired Democratic leader Richard Gephardt have both taken a data-based approach to dissuade centrists like Joe Manchin from joining a Unity ticket.
Who’s right?
Can No Labels represent the Third Mountain of voters?
The fundamental promise of the Unity ticket concept is that the needs of unrepresented voters, who are neither Democrats or Republicans, are unmet by the current electoral system. No Labels argues that a Unity ticket can meet those needs – and maybe even elect a president.
They argue further that in this year‘s election, which is heading inevitably towards a rematch of President Joe Biden versus former president Donald Trump, voters are dissatisfied with their choices. President Biden‘s approval ratings are low. Donald Trump is an unpopular former president and an insurrectionist who makes statements that call into risk the foundations of democracy itself. (Note that many No Labels adherents acknowledge that re-electing a bad president and destroying democracy are not equally bad choices.)
You might think that this sounds a lot like my thesis that a third mountain of unrepresented voters is unrepresented. (Go read that please!) However, the question is how to achieve that representation. In this regard, the No Labels plan faces several major obstacles.
The third-choice paradox
First, there have to be enough voters who want a third choice to make this viable strategy. Nominally, that is true. Self-reported independents have been more numerous than Democrats and Republicans in most years since 1991. This year independents hit an all-time high, 43 percent.
However, third-party candidates do not do well in American politics. Since 1900, the top eight third-party candidates averaged just 11.8 percent of the vote and 23 electoral votes (this year, 270 electoral votes are needed to win). The high point was Theodore Roosevelt, a Republican former president who started the Bull Moose Party. He ended up with 27.4 percent of the popular vote and 88 electoral votes, and threw the 1912 election to the Democrats. The most recent specimen, Ralph Nader in 2000, got 2.7 percent, zero electoral votes, and arguably threw the election to Republican George W. Bush.
How can both of these facts - a mountain of independents on the one hand, and third-party failure on the other - both be true? The answer lies in two parts: (a) the rules of how we select a president, and (b) what I call “voter geometry.”
To understand this it is helpful to have some diagrams to illustrate how rules and geometry collide.
Imagine that third mountain, off to the side. People in that third mountain are different from the other two mountains - in tone, or by issues, it doesn’t matter how. We can imagine them sitting apart, at a distance.
But in our election system they only get to make one choice, a rule that is called first-past-the-post or plurality voting. With two major parties in the running, naturally they will pick the alternative that is closer to them.
In terms of geometry, we would call this projecting high-dimensional complexity onto one dimension. When voters view everything as falling on one axis, either the Democrat or the Republican will usually look better.
No Labels would argue that we need to find someone who hits that Third Mountain. At the moment, the closest thing is Robert Kennedy Jr., anti-vaccine activist and possessor of an amazing last name. Certainly anti-vaccination sentiments have appeal across the left-to-right spectrum. So that’s an option.
Hating both the players and the game
The second, larger problem for No Labels is that the plurality voting rule in America does not reward minor candidates. Based on the diagrams above, you might imagine that a third-party candidate tends to hurt whoever is closer to them. Indeed that is so, as can be shown by computer simulation.
At the Electoral Innovation Lab, we used real voting data to figure out how much voters fall into the two mountains, and how many are in between. I then used that information to simulate voter behavior in a hypothetical three-candidate presidential race.
The real voting data come from FairVote and the Ranked Choice Voting Research Center, which maintain a database of over over 400 ranked-choice elections. In these elections, voter rankings provide a rich source of information. We can use them to figure out how candidates are seen by the voting public as a whole. As it turns out, voters mostly see candidates as falling along a single dimension, and voters are arranged on the same dimension. Averaged across many dozens of elections in each category, the voters look like this:
In each of these “voter spectra,” voters and candidates fall along a single dimension. Furthermore, in partisan elections (red), the voters are arranged into two peaks, which correspond to liberals and conservatives. The vast majority of voters have preferences that fit on this left-right axis. The two peaks are highly reminiscent of the two mountains in my graphic.
Now let’s see what happens in a simulated election, where a plurality vote is enough to win a state, as happens in the Electoral College. A closely-matched two-candidate election looks like this:
On the left is how voters break down on their preferences between Biden and Trump. On the right is a graph showing how much of the vote it takes to win the election. Here, Biden and Trump each need typically need 50% of the vote. as one would expect for a two candidate race. In 10,000 simulations, they each win about 5,000 times.
Now imagine a Unity ticket that sits in the middle, between Biden and Trump. Polling evidence suggests that Trump voters are highly loyal - he’s been in the 45-47% range ever since 2016. As of right now, Biden is running even or slightly behind him in polls, suggesting that support for Biden is someone softer. Therefore, let’s put the Unity ticket slightly closer to Biden.
Now the outcome looks like this:
Here, the simulations turn out quite differently. Now, Donald Trump wins 8,000 of the simulations while President Biden wins only 2,000 of the simulations. The Unity ticket never wins. Why is this?
The answer is found in the right-hand graph. Biden still needs 50 percent of the vote to win. But Trump‘s necessary vote share has dropped to 46 percent. That’s because the Unity ticket candidate draws from the candidate he or she is closer to, even if they are only a little closer. The in-between voters, shaded in gray, are just enough to create a spoiler effect. The Unity ticket candidate has spoiled the chances of Joe Biden, who would otherwise be more acceptable to a majority of voters than Trump.
In short, when it comes to electing a president under current rules, the No Labels approach is doomed. The reason: In a plurality system, a middle candidate undermines the opponent who is closer to his/her ideology.
However, there is a way for No Labels to have better success.
Another path for No Labels
Now consider an alternative scenario. In the analysis of voting records, we found that voters behaved differently in regional races. They were more evenly distributed along the left-right axis - in other words, less polarized. What happens when we run an election under those conditions?
Now we see a very different result. In this case, the Republican and Democratic candidate each win 4,200 times – and the third-party candidate wins 1,600 times. In this scenario, a third-party candidate is now viable.
Furthermore, all three candidates can potentially win with less than 50 percent of the vote. Indeed, in these particular simulations the No Labels candidate typically wins with 36 to 38 percent of the vote.
As it turns out, there are numerous offices that are less polarized than partisan offices: board, local executive, and local representative offices.
If the goal of the No Labels movement is to build a centrist alternative, their path goes through local elections. They get to build a vital center - without undermining the major party they dislike less. It is not unlike the approach taken by Emmanuel Macron, who started a new party which he took all the way to the presidency of France.
Yet another path: reprogram the game
There is yet another route for No Labels to succeed, which is to change the voting rule itself. Call it the Kobayashi Maru scenario.
All the simulations above used first-past-the-post voting. But what if we had a ranked-choice system? In such a system, voters may cast a vote for a minor-party candidate, then reassign their vote to a more likely winner if their first choice loses. This allows them to express their preference without ultimately throwing their vote away.
Such a system is in place in dozens of localities around the nation. Indeed, that is where I got the ranked-choice voting data to calculate the voter spectra in the first place. If No Labels wants to make a long-term difference, they can advocate for such rules in local races, which they can then use to build a better politics from the bottom up.
Such a strategy is not a pipe dream. In Alaska, a version of ranked-choice voting called Top Four has been enacted. Already under this rule, there are indications that vying for second-choice votes can sand the edges off a rough campaign. After just two elections under the Top-Four rules, famous bomb-thrower Sarah Palin (R) has started to see the virtues of being friendly with her opponents. Here she is with opponents Mary Peltola (D) and Nick Begich (R).
In the weeks and months ahead here at the Electoral Innovation Lab, we’re diving into the question of how long it takes to achieve a Peltola-Palin-Begich-like effect. We’re also asking if it depends on the type of election. Our research will help democracy-repair organizations around the nation decide where to target their energies. Based on their ideals, No Labels should be one of those organizations.
I thank Mahshad Habibpourparizi, Eric Simpson, and Keena Lipsitz for collaboration in the analysis of ranked-choice voting records.
Very interesting, and confirms my gut suspicion. But writing about No Labels doesn't mean you should ignore your high school math teacher by not labeling your axes.
The attribution of "good intentions" to No Labels is unsupported by evidence, and is a highly controversial stance.